Zsh Mailing List Archive
Messages sorted by: Reverse Date, Date, Thread, Author

[BUG] PCRE match hangs when matching U+00A0



After enabling the shell option «rematchpcre», a zsh (ver. 5.2) match
freezes when the matched string contains the character «NO-BREAK SPACE»
(U+00A0).

This may be reproduced by running the following:

    setopt rematchpcre
    s=$(echo -en \\ua0)
    [[ $s =~ . ]]

This occurred while using the en_US.utf-8 locale.

The hang seems to arise from an infinite loop at pcre.c:202
(zpcre_get_substrings). At least gdb never showed zsh returning from
this stack frame. A typical backtrace after interrupting the loop is:

	#0  mb_metacharlenconv_r (s=0x716542 "\240b", wcp=0x0, mbsp=0x6d8f28 <mb_shiftstate>) at utils.c:4988
	#1  0x00000000004a7143 in mb_metacharlenconv (s=0x716542 "\240b", wcp=0x0) at utils.c:5031
	#2  0x00007ffff5eec71d in zpcre_get_substrings (arg=0x716540 "a b", ovec=0x7165e0, ret=1, matchvar=0x7ffff5eed0df "MATCH", substravar=0x7ffff5eed0e5 "match", want_offset_pair=0, matchedinarr=0, want_begin_end=1) at pcre.c:204
	#3  0x00007ffff5eecf4b in cond_pcre_match (a=0x7ffff7ff2778, id=0) at pcre.c:389
	#4  0x0000000000426574 in evalcond (state=0x7fffffffdab0, fromtest=0x0) at cond.c:155
	#5  0x0000000000433d5a in execcond (state=0x7fffffffdab0, do_exec=0) at exec.c:4530
	#6  0x000000000042a42d in execsimple (state=0x7fffffffdab0) at exec.c:1131
	#7  0x000000000042a7a2 in execlist (state=0x7fffffffdab0, dont_change_job=0, exiting=0) at exec.c:1255
	#8  0x000000000042a14a in execode (p=0x7ffff7ff25e8, dont_change_job=0, exiting=0, context=0x4b17b7 "toplevel") at exec.c:1075
	#9  0x000000000044b8a9 in loop (toplevel=1, justonce=0) at init.c:208
	#10 0x000000000044f144 in zsh_main (argc=2, argv=0x7fffffffdd88) at init.c:1680
	#11 0x000000000040f366 in main (argc=2, argv=0x7fffffffdd88) at ./main.c:93

While stepping through, I’ve also noticed that mb_metacharlenconv_r hits:

	4991 	return 0;		/* Probably shouldn't happen */

Best regards,

Sebastian Reuße



Messages sorted by: Reverse Date, Date, Thread, Author